Got an error that I could find for a while. It was when parsing a lot of
files.
Turn out it was the Thread Pool or something because if I run on 1 core,
it's ok.
Different thing in this commit, I just want to freeze a state that
benchmark work.
Now I flush only when the file is full and I check the the currently
used file if it is big enough.
So I dont get stat of all files and flush everytime like before
So then I can parse the file again, create a map of UUID sub json and
just iterate over the first json looking for UUID and update with new
data from map
